Predictive learning

A subset of Machine learning where an agent tries to build a model of its environment, by predicting what will happen when it takes various actions. It uses this knowledge as planning operators. Predictive learning is one way forward toward bootstrapping knowledge. It must start with some pre-existing knowledge, but its major goal is to build more knowledge on top of it. Gary Drescher's Schema mechanism was a seminal use of predictive learning to try to re-create Piaget's stages of learning in a machine.